Statistikklubben LAB: Sales Prediction with XGBoost (Sthlm)

Statistikklubben LAB: Sales Prediction with XGBoost

NOTE: Spots are limited. If you’re not able to join this lab, you’re very welcome to do the project on your own and then join the follow-up online session for questions and discussion a week later (let me know if you want to join this second event).

In this hands-on lab, we will work with the same basic setup, built around two core components:
Model: XGBoost (or your model of choice, see below)
Task: Predicting sales outcomes

Level
Basic knowledge of Python and some familiarity with machine learning or statistical modeling is recommended.

XGBoost
XGBoost is a natural choice for this kind of project. It’s powerful, flexible, well‑documented and widely used in real‑world forecasting tasks, yet still accessible enough that you don’t need deep domain expertise to get started. The sales‑prediction problem itself strikes a nice balance: realistic, but not so specialized that it requires industry knowledge.

Digging deeper
If you’re already comfortable with XGBoost, that actually opens up more room to focus on other fun parts:

• Data preprocessing
• Exploratory analysis
• Feature engineering
• Handling missing values and outliers
• Experimenting with validation strategies

In other words, all the fun parts where statistical thinking and creativity really matter.

Other models
For those who already feel at home with XGBoost, there’s room to explore alternatives. If you want to try a different model (neural networks, random forests, linear models, or something more exotic) you’re absolutely encouraged to do so. Comparing results across different approaches can give us a richer understanding of model behavior and might help us see why certain methods perform better in some situations than others.

The goal is not just to build a model, but to learn from each other, compare methodologies, and deepen our understanding of predictive modeling as a whole.

➡️ One week before the lab, I will post t